Specilized API for IO #7
Conversation
| @@ -20,7 +20,7 @@ import cats.implicits._ | |||
| import munit.CatsEffectSuite | |||
|
|
|||
| class CatchTest extends CatsEffectSuite { | |||
| val C = Catch.ioCatch.unsafeRunSync() | |||
| val C = Catch.catchForIO.unsafeRunSync() | |||
There was a problem hiding this comment.
Renamed this to avoid confusion between ioCatch and ioCatching
There was a problem hiding this comment.
Reverted as we're using IOCatch[E] now
|
|
||
| object IOCatch { | ||
|
|
||
| def ioCatching[E]: IOCatchPartiallyApplied[E] = |
There was a problem hiding this comment.
Alternatively can be named catchingIO. Thoughts?
There was a problem hiding this comment.
Do you have an opinion on apply?
IOCatch[Error] { h =>
// ...
}There was a problem hiding this comment.
Good point - that didn't cross my mind 😄 Done
| for { | ||
| res <- ioCatching[String](h => | ||
| IO.uncancelable(poll => poll(h.raise("oops"))) | ||
| ) |
There was a problem hiding this comment.
This one is interesting because I couldn't get the RaisedInUncancellable exception to be thrown unless I remove the race in
Perhaps there's a race condition and somehow the outer cancellation always wins? 🤔
Either way, we will need to use raiseError to abort the execution anyway so nothing changes.
There was a problem hiding this comment.
Yeah I just tried prom.get <* F.sleep(1.second) and it raised the error.
I think it is very hard to have it throw consistently without being able to query cats-effect about the "current cancelation" context.
Unfortunately I think that also means that the test "Can raise in an uncancellable region" might be flaky 😿
There was a problem hiding this comment.
I guess we can make both cases pass for now since it's race-y 🤷♂️ Will need to document this behaviour
|
Happy to add other missing test scenarios if you can think of anything else 🙂 |
| for { | ||
| res <- ioCatching[String](h => | ||
| IO.uncancelable(poll => poll(h.raise("oops"))) | ||
| ) |
There was a problem hiding this comment.
Yeah I just tried prom.get <* F.sleep(1.second) and it raised the error.
I think it is very hard to have it throw consistently without being able to query cats-effect about the "current cancelation" context.
Unfortunately I think that also means that the test "Can raise in an uncancellable region" might be flaky 😿
|
Some of the |
|
May I merge @jatcwang ? |
|
Yes go for it thanks @ValdemarGr! |
Closes #6